In 1986, an educational researcher named Keith Stanovich coined a term that has since been borrowed across half a dozen fields. He called it the Matthew effect in education. The name comes from the Gospel of Matthew — to those who have, more shall be given — and Stanovich was using it to describe something specific that happens to children learning to read.
Two children start Year 1. One reads slightly more fluently than the other. The difference at the start is small — perhaps the equivalent of a few months of development. The slightly more fluent reader finds reading slightly easier. Because reading is slightly easier, she does slightly more of it. Because she does more of it, her vocabulary expands slightly faster. Because her vocabulary is slightly larger, she finds reading at the next level slightly easier. The gap widens by a small amount.
By Year 2, the difference is no longer the equivalent of a few months. By Year 4, it might be the equivalent of two years. By Year 6, the two children — who started with what looked like a minor difference — are reading at substantially different levels, accessing different kinds of texts, building different kinds of vocabulary, developing different kinds of relationships with school itself. The gap that began as a small variance has become an order-of-magnitude difference, and almost every step along the way was the natural consequence of the step before it.
What’s striking about this is what it doesn’t require. No bad teaching. No parental neglect. No flaw in the slower-reading child. The mechanism doesn’t depend on anyone making a wrong choice. It depends only on the structural fact that reading is a skill where each year’s level shapes the next year’s practice, which shapes the next year’s level. Small initial differences compound, through ordinary functioning of the system, into large eventual ones.
This pattern — small starting differences compounding into large ending ones through the structural feedback of normal life — is one of the most important and least-noticed features of the world. It produces outcomes that look like talent or luck or merit. The mechanism is more specific and more traceable than any of those words suggest.
A small shift in the question you ask
When you observe outcomes that vary substantially across people — different levels of skill, success, prosperity, status — most people’s instinct is to ask why is that person where they are? The question implicitly looks for explanations in the individual. Talent. Effort. Choice. Luck.
A better first question is: what mechanism turned a small initial difference into the large eventual one I’m seeing now?
The shift is from looking at the endpoint to looking at the process that produced it. The endpoint is the visible thing. The process is usually invisible — operating across years, in small increments, through ordinary functioning that nobody particularly notices. But the process is what produced the gap. Without seeing the process, the gap looks unexplained, and the explanations people reach for tend to fall into two unhelpful categories: the gap is the result of inherent differences (which gives up on explanation), or it’s the result of someone’s wrongdoing (which often isn’t true).
The structural answer — that small differences compound through specific mechanisms — is more accurate, more explanatory, and more useful than either default. It also applies symmetrically. The same compounding that turns small early advantages into large later ones turns small early disadvantages into large later ones. And small early efforts compound the same way. The mechanism doesn’t care about the moral valence of what’s being compounded. It just compounds whatever’s there.
The instinct in action
Three cases where small starting differences turn into large eventual ones through specific, traceable mechanisms.
How reading creates more reading
The Stanovich pattern is worth working through carefully because it shows the mechanism in pure form. A child whose reading is slightly more fluent encounters less friction in books. Less friction means they read for slightly longer at a stretch, encounter slightly more words per session, and feel slightly more pleasure than effort. Pleasure produces repetition; repetition produces practice; practice produces more fluency. Meanwhile, the child whose reading is slightly less fluent encounters slightly more friction. Friction produces fatigue, fatigue produces avoidance, avoidance produces less practice — and the gap widens.
This isn’t a story about ability or effort in any simple sense. The slightly less fluent reader isn’t lazy. They’re operating in a structural feedback loop that produces less practice from a slightly less fluent starting point. The slightly more fluent reader isn’t necessarily smarter. They’re operating in the same loop running in the favourable direction.
The mechanism is not unique to reading. Anything that meets these conditions — current level shapes practice, practice shapes future level, repeated over enough cycles — will compound small differences into large ones. Musical instruments. Sports. Social skills. Almost any domain where you get better through doing more of it.
How professional networks compound their members’ opportunities
A young person takes their first professional job. Two graduates of the same university, with similar grades and similar interviews, end up at different starting firms. One ends up at a well-regarded firm where many junior employees later become senior people in the industry. The other ends up at a smaller, less-networked firm, doing similar work for similar pay.
Twenty years later, the differences between these two careers will probably be substantial — not because of differences in ability or effort, but because of who they spent their early career standing next to. The graduate at the well-regarded firm worked alongside people who, over twenty years, scattered across the industry into senior roles. They share weak professional ties with hundreds of senior people who can introduce them to opportunities, vouch for them when their name comes up, mention them when a position opens. The graduate at the smaller firm has fewer of those ties — not because they did anything wrong, but because the network they happened to be embedded in produced fewer of them.
The American sociologist Mark Granovetter described this dynamic in 1973 in a paper called The Strength of Weak Ties. Strong ties — close friends, immediate colleagues — are emotionally important but produce limited opportunity flow because the people you’re closest to mostly know what you know. Weak ties — distant acquaintances, people you’ve worked with briefly — produce most of the opportunities that change a career, because they’re embedded in different information flows than your own.
A starting position determines what weak ties you accumulate. Weak ties shape what opportunities you hear about. Opportunities shape your next position, which shapes the weak ties you accumulate next. Over twenty years, two careers diverge substantially. Neither person has done anything wrong; the structural feedback of professional networks has compounded a small initial difference into a much larger one. Some people land in well-connected starting positions through deliberate effort, some through luck, some through resources their families had, some through advice they happened to receive. Once the starting position is set, the compounding does much of the rest.
How neighbourhoods compound advantages across generations
The most carefully documented version of this mechanism comes from research by the American economist Raj Chetty and his collaborators at Harvard, who have spent over a decade tracking how children’s outcomes vary by where they grew up.
The findings are striking and largely apolitical in their framing. Children who grow up in neighbourhoods with certain features — stronger schools, lower crime rates, more professional adults around them — have higher chances of attending university, of accessing certain kinds of jobs, and of being able to afford certain kinds of neighbourhoods themselves as adults. Even moving the same child from one type of neighbourhood to another in early childhood appears to change their adult outcomes measurably.
The mechanism is the same compounding we’ve already seen, but operating across generations rather than years. A neighbourhood with better schools produces children who attend university at higher rates. University-educated adults earn more, on average, and tend to choose neighbourhoods with similar features for their own children. Those children then have access to the same advantages their parents had, plus whatever their parents accumulated. Across two or three generations, the gap between two starting neighbourhoods grows substantially — without anyone in the chain having made bad decisions.
The same mechanism runs in the opposite direction. A child growing up in a neighbourhood with weaker schools and fewer professional adults has lower probabilities of attending university, of accessing professional careers, and of being able to afford different neighbourhoods for their own children. None of this is destiny — individual children obviously do break out of the pattern, and most don’t have lives determined by their postcode. But the structural feedback shapes the probabilities.
Chetty’s work is notable in part for how carefully it avoids prescribing political conclusions. The research describes the mechanism. What to do about it — whether to address it, how, through which institutions, with what costs — is a separate set of questions the research itself doesn’t try to answer. But seeing the mechanism is what makes the questions sensible to ask.
Where this way of seeing comes from
The intellectual tradition behind compounding-mechanisms thinking has multiple roots. The Matthew-effect framing comes originally from the sociologist of science Robert K. Merton, who in 1968 used it to describe how scientific recognition concentrates on already-recognised researchers. Stanovich’s 1986 application to reading extended the idea into education. In economics, the formal study of cumulative advantage has a long history — Pareto’s observations in the 1890s on wealth distributions, Herbert Simon’s work on related dynamics in the 1950s, and contemporary research on inequality dynamics by economists including Thomas Piketty.
The systems-thinking version of this tradition focuses less on the mathematical patterns and more on the underlying mechanism: a feedback loop where the current state shapes the next-step input, which shapes the next state. Donella Meadows called these reinforcing loops in Thinking in Systems (2008), and identified them as the source of most patterns in human systems where small differences become large ones. The loops are domain-general — they show up in skill development, in financial accumulation, in scientific reputation, in athletic performance, in friendships. Wherever current level shapes future level through ordinary practice, compounding will eventually produce substantial spread.
Where this has limits
Three caveats matter here, and matter more than usual because the topic is one where careful framing is the difference between insight and ideology.
First, compounding mechanisms describe what’s happening structurally — they don’t describe what should be done about it. The mechanisms are present in any system where current state shapes future practice. Whether the resulting spread is just, unjust, addressable, or inevitable is a separate set of questions that the structural analysis doesn’t answer. Different political traditions answer those further questions differently, with different evidence and different values. The structural perception is a starting point, not a conclusion.
Second, compounding doesn’t eliminate individual agency. The Year 1 reader who falls behind isn’t fated to a particular educational future; the graduate at the smaller firm isn’t fated to a particular career. People make choices that matter, sometimes substantially, even within structural patterns that shape probabilities. The structural mechanism shapes the field on which choices are made, but it doesn’t make the choices for anyone. Treating compounding as destiny is both empirically wrong and corrosive to the sense of agency students need to live their lives well.
Third, compounding works in both directions. The same mechanism that produces large gaps from small disadvantages also produces large gains from small advantages — and produces large skill from small efforts when those efforts are sustained. A student who practises a small amount each day for ten years ends up substantially better than someone who practises sporadically for the same total hours. The mechanism is morally neutral; it amplifies whatever is being repeated, in whichever direction.
The question that remains
The deepest thing this strand of systems thinking teaches is probably this. Most of the variation you see in the world — in skill levels, in successes, in prosperity, in fluency, in relationships, in everything accumulated through repeated practice or repeated transmission — was produced by mechanisms that were running before anyone noticed. The endpoint looks like a fact about the people involved. The process that produced it was structural, mostly invisible, and operating on a timescale long enough that it became almost impossible to see.
Once you can see the process, the endpoint looks different. Not less real, not less consequential, not less in need of being responded to in whatever way the situation calls for. Just different. You can no longer treat it as the simple result of how the people involved happened to be. You see, instead, the slow accumulation of small differences through structures that have been quietly working all along.
The question worth carrying:
For the next week, every time you notice an uneven outcome — different skill levels among classmates, different career trajectories among adults you know, different prosperity across neighbourhoods you pass through — pause and ask the second question. What mechanism turned small initial differences into the large eventual ones I’m seeing now? Notice that the answer almost always involves something that’s been running for years, in small increments, through ordinary functioning that nobody particularly noticed.
Key research referenced: Robert K. Merton, The Matthew Effect in Science (1968); Keith Stanovich, Matthew Effects in Reading (1986); Mark Granovetter, The Strength of Weak Ties (1973); Raj Chetty’s neighbourhood-effects research at Harvard’s Opportunity Insights project; Donella Meadows on reinforcing loops in Thinking in Systems (2008); Pareto distributions and cumulative-advantage research in economics and sociology.